BLXBench
Code · 2026-04-24
BLXBench evaluates AI coding models on software-engineering tasks such as UI generation, architecture, debugging, and code writing. Its public leaderboard reports an overall score plus pass rate, latency, token throughput, cost, task-domain, and difficulty breakdowns.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.5 | 77.9 |
| GPT-5.3-Codex | 77.7 |
| Qwen3.7-Max | 75.3 |
| Kimi K2.6 | 74.8 |
| MiniMax M3 | 74.7 |
| DeepSeek-V4-Pro | 73.8 |
| MiMo-V2.5 | 73.8 |
| MiMo-V2.5-Pro | 73.1 |
| Grok Build 0.1 | 72.8 |
| Nemotron 3 Super 120B A12B | 72.3 |
| GLM-5.1 | 72.1 |
| DeepSeek-V4-Flash | 71.2 |
| GLM-5.2 | 70.8 |
| Mistral Medium 3.5 | 70.3 |
| Qwen3.7-Plus | 69.9 |
| Opus 4.8 | 69.7 |
| Qwen3.6-Flash-2026-04-16 | 69.3 |
| CoBuddy | 68.7 |
| Opus 4.7 | 68.1 |
| Nemotron 3 Nano 30B A3B | 68.0 |
| Mistral Small 4 | 67.8 |
| Grok 4.3 | 67.6 |
| Granite 4.1 8B | 67.1 |
| North Mini Code | 66.3 |
| Kimi K2.7 Code | 65.7 |
| Ring 2.6 1T | 65.3 |
| Claude Fable 5 | 64.5 |
| Gemini 3.1 Flash-Lite | 64.2 |
| Nemotron 3 Nano Omni 30B A3B | 57.9 |
| Step 3.7 Flash | 55.2 |
| MiniMax M2.7 | 51.0 |
| Gemini 3.5 Flash | 41.5 |